Computational and data driven molecular material design assisted by low scaling quantum mechanics calculations and...

Area of Science:

• Computational chemistry
• Materials science
• Quantum mechanics

Background:

• Quantum mechanics (QM) methods like DFT, TDDFT, and WFT are crucial for predicting molecular and optoelectronic properties.
• High computational costs of QM methods limit their application to large systems, posing a challenge for molecular material design.

Purpose of the Study:

• To review low-scaling quantum mechanics (QM) approaches and machine learning (ML) techniques for computational molecular material design.
• To illustrate the application of these methods to various complex molecular systems and properties.

Main Methods:

• Low-scaling ground-state and excited-state QM methods.
• Machine learning (ML) algorithms including deep learning and on-line learning.
• Modified force fields with polarization models and variable electrostatic parameters.

Main Results:

• Demonstrated applications of low-scaling QM to long oligomers, supramolecular complexes, and stimuli-responsive materials.
• Utilized ML algorithms to predict molecular energies, forces, electronic structure, and optical/electrical properties.
• Introduced modified force fields with polarization for enhanced electrostatic modeling.

Conclusions:

• Low-scaling QM and ML techniques significantly reduce computational costs for molecular material design.
• Future directions include combining low-scaling algorithms with periodic boundary conditions and ML for periodic functional materials.